Lesson 51 — Testing, Sharing and Maintaining Sentinel Workbooks
A workbook is only operationally useful when analysts can trust its data, open its evidence and access it with the correct permissions. Learn to validate, publish and maintain your SOC dashboard safely.

What you will learn
Turn your Lesson 50 dashboard into a supportable SOC asset.
Lesson 51 — From dashboard to dependable SOC tool
In Lesson 50, you built a SOC Operations Dashboard. Now Agent Foskett has a more difficult question: will it still work on Monday morning, for the next analyst, after the data changes?
Testing, controlled sharing and maintenance turn a useful workbook into an operational resource.
Learning objectives
- Validate workbook data against source records.
- Test parameters, links and empty-data behaviour.
- Separate workbook access from data access.
- Check query performance and refresh expectations.
- Publish with an owner, change log and rollback plan.
- Run a practical handover acceptance test.
The operational lifecycle
Prepare the lab
- Open your SOC Operations Dashboard — Lab workbook from Lesson 50.
- Save a separate test copy named SOC Operations Dashboard — Validation.
- Record the workspace, resource group, subscription and workbook owner.
- Choose a test period with known incidents and a period with few or no events.
- Identify a second test account with the intended analyst permissions.
Do not test destructive incident actions or expose production identities in screenshots.
Step 1 — Create a test matrix
| Test | Pass condition | Evidence to capture |
|---|---|---|
| KPI totals | Open, high and unassigned counts reconcile | Source query and screenshot |
| Triage queue | Latest incident state shown; row limit disclosed | Incident IDs and timestamps |
| Parameters | Date/workspace selections change intended panels | Before/after results |
| Navigation | Links open the correct permitted record | Target incident |
| Empty data | No misleading zero or false success message | Empty-period screenshot |
| Permissions | Analyst sees only authorised content | Role and test result |
| Performance | Loads within your agreed operational target | Measured elapsed time |
Step 2 — Validate latest incident state
Run a source query independently of the workbook. The latest incident record matters because SecurityIncident may contain multiple versions of the same incident.
- 1
- 2
- 3
- 4
- 5
- 6
- 7
SecurityIncident
| summarize arg_max(TimeGenerated, *)
by IncidentNumber
| where Status != "Closed"
| summarize OpenIncidents = count(),
HighSeverity = countif(Severity == "High")Compare these values with the dashboard. For multi-workspace reports, include a workspace-aware incident key. The latest row means the latest available in the queried data, not necessarily the current state if ingestion is delayed.
Step 3 — Check duplicate and missing identifiers
Check whether incident updates exist and whether your chosen incident key is populated. Multiple records per incident are expected; blank identifiers warrant investigation.
- 1
- 2
- 3
- 4
- 5
- 6
- 7
SecurityIncident
| summarize Records = count(),
LatestRecord = max(TimeGenerated)
by IncidentNumber
| order by Records desc
| take 20This query is diagnostic. It does not prove data completeness and should not be used as the dashboard’s current-state count.
Step 4 — Test time parameters
- Set the workbook to a 24-hour period and note the creation-trend count.
- Change to seven days; confirm the trend changes as expected.
- Check that the current open-incident count is not unintentionally restricted by the trend parameter.
- Change any workspace or severity parameters and inspect every dependent panel.
- Reload the workbook and verify defaults.
A control that appears on screen but does not affect its intended query is a defect.
Step 5 — Test navigation and empty states
- Open an incident from the triage table and verify the identifier.
- Confirm links work for the analyst role, not only for the author.
- Choose an empty time window: display No data for selected period rather than implying the environment is safe.
- Check missing-table and permission-error messages.
- Confirm row caps and time scopes are labelled.
Step 6 — Verify telemetry freshness
A workbook can render successfully while its underlying data is stale. Display freshness for each required source and investigate unexpected gaps.
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
union isfuzzy=true
(SecurityIncident
| summarize Latest = max(TimeGenerated)
| extend Source = "SecurityIncident"),
(SigninLogs
| summarize Latest = max(TimeGenerated)
| extend Source = "SigninLogs")
| project Source, Latest,
MinutesSinceLatest =
datetime_diff("minute", now(), Latest)Run each source independently if one is absent. Define freshness thresholds based on connector expectations; a recent timestamp alone does not prove complete ingestion.
Step 7 — Understand workbook permissions
| Permission layer | What to verify |
|---|---|
| Workbook resource | The analyst can open the shared workbook resource. |
| Log Analytics workspace | The analyst can read the tables queried by the workbook. |
| Microsoft Sentinel incidents | The analyst can open incident evidence appropriate to their role. |
| Connected resources | Cross-workspace or cross-resource queries have the necessary access. |
| Edit rights | Only authorised maintainers can change production content. |
Workbook access does not automatically grant access to Log Analytics data or incidents. Assign the least-privileged built-in roles that satisfy your environment, and test with a real non-admin account.
Step 8 — Share safely
- Confirm the workbook is saved as a shared Azure workbook resource in the intended subscription and resource group.
- Open the workbook’s resource and access-control settings.
- Grant the analyst access needed to view the workbook and its underlying data, following your organisation’s RBAC model.
- Share the workbook resource link with the analyst.
- Sign in as the analyst and test the dashboard and incident drill-down.
Do not confuse sharing a link or exported workbook template with granting access to the queried data. Portal controls and role assignments vary by tenant.
Step 9 — Test query performance
Measure representative loads with realistic incident volumes and selected time ranges. Investigate slow panels before increasing automatic refresh frequency.
- Filter event tables to the required period when semantically appropriate.
- Keep current-state incident logic separate from historical event trends.
- Use
projectto return only fields needed for a visual. - Avoid unnecessary repeated expensive queries and oversized detail tables.
- Check workspace query diagnostics and costs where available.
Do not apply a narrow TimeGenerated filter before latest-state aggregation if it would hide older open incidents.
Step 10 — Check refresh behaviour
- Document the expected refresh interval and data latency.
- Check what happens when a workbook remains open for an entire shift.
- Verify timestamps update when expected.
- Balance operational needs against query and ingestion cost.
- Display a visible last-reviewed note, not an unverified “live” label.
Step 11 — Version control and rollback
| Release item | Example entry |
|---|---|
| Workbook name | SOC Operations Dashboard |
| Version | 1.0.0 |
| Owner | SOC Operations Lead |
| Environment | Validation / Production |
| Changes | Added incident freshness checks |
| Test status | Passed with analyst account |
| Rollback | Restore last approved workbook definition |
Export or otherwise retain the previous approved workbook definition according to your deployment process. Store copies in an access-controlled location and avoid including secrets or sensitive data in source control.
Step 12 — Maintenance runbook
- Review broken queries, permission failures and unexpected blank panels.
- Check connector status, data freshness and schema changes.
- Review query latency and usage cost.
- Retest incident links after portal or workbook changes.
- Review owner, access, documentation and change history.
- After any update, repeat the acceptance tests before production release.
Step 13 — Troubleshooting guide
| Symptom | Likely checks | Safe next action |
|---|---|---|
| Workbook opens but panels are blank | Workspace selection, time window, RBAC, source tables | Test source query directly |
| KPI disagrees with incident list | Versioned records, filters, scope, ingestion delay | Reconcile incident IDs |
| Analyst sees access denied | Workbook versus workspace permissions | Review role assignments |
| Workbook loads slowly | Broad scans, repeated queries, large tables | Measure and optimise |
| Link opens wrong incident | Incorrect parameter or identifier mapping | Verify with known lab incident |
| Data appears stale | Connector or ingestion delay, refresh settings | Check source timestamps |
Agent Foskett case file — The perfect dashboard nobody could open
Monday, 08:05. The SOC dashboard looked magnificent on the engineer’s screen. Twelve KPI cards, five charts and a reassuringly green status panel.
Then the incoming analyst opened the link.
“Access denied.”
Agent Foskett examined the evidence. The workbook was shared. The workspace was not. The dashboard was not broken — its access design was incomplete.
“Excellent,” he said. “We’ve successfully secured the dashboard from the people who need it.”
The team corrected the permissions, tested with the analyst account and recorded the release. This time the green status meant something.
Hands-on exercise — Production readiness review
- Create a validation copy of the Lesson 50 workbook.
- Complete the test matrix for KPIs, trends, navigation and empty states.
- Independently reconcile open incident totals.
- Check freshness for
SecurityIncidentand one supporting data source. - Measure dashboard load time with two time windows.
- Open the workbook with a non-admin analyst account.
- Record at least one simulated failure and its resolution.
- Produce a release note and rollback reference.
Knowledge check
- Why can counting all
SecurityIncidentrows inflate totals? - Does permission to open a workbook automatically grant access to its workspace?
- What should an empty chart say when the selected period has no records?
- Why should current backlog counts be tested separately from time-filtered trends?
- What evidence would you keep before releasing a workbook update?
Check your answers: incident updates create multiple rows; no; clearly state no data for the selected scope; old open incidents may still need action; preserve test results, approvals, version details and rollback material.
Release checklist
- ✓ Source totals reconcile with dashboard visuals.
- ✓ Time parameters and workspace scope are labelled and tested.
- ✓ Empty states and missing sources are handled honestly.
- ✓ Incident links lead to the correct evidence.
- ✓ Analyst permissions work without administrator privileges.
- ✓ Performance and refresh expectations are documented.
- ✓ Owner, version, change log and rollback copy exist.
- ✓ No sensitive production data appears in public material.
Agent Foskett takeaway
A dashboard is not ready because it looks good. It is ready when the next analyst can trust it, use it and recover it when something changes.
The logs already knew. This time, so did the change record.
